The Canadian Intelligent Systems Challenge is a series of annually held Artificial Intelligence competitions open to Canadian graduate, undergraduate and high school students. The competition is a joint initiative by the Canadian Artificial Intelligence Association and Precarn Incorporated.
Contents
Mission
The main objectives of the competition are:
Process
Each year, proposals for challenge problems are elicited from Canadian companies and the winning challenge problem is selected by a group of experts from Canadian academia and industry who also organize the competition. The eligibility of a challenge problem is that it be something that a company would like to see handled by a software program, that It must be possible to define the challenge problem in a precise manner and the problem should not be trivial.
Each year, cash prizes will be awarded in three categories:
There will also be a $10,000 cash prize to the team that produces the overall best solution to the challenge problem.
Intelligent Systems Challenge 2009
The Challenge problem for 2009, titled Protecting Canada's Coastal Border, was proposed by MacDonald Dettwiler and Associates (MDA), the aerospace and financial products company that built the Canadarm, Radarsat-2 and image processing systems used for Google Earth. The organizing committee consisted of faculty members of University of Waterloo, Simon Fraser University and Dalhousie University.
Problem Statement
Various Canadian government agencies are responsible for protecting Canada, asserting sovereignty, and enforcing the law in maritime approaches to Canada. They thus continually watch for ships that are behaving strangely, based on data from surveillance flights, voluntary reports from the ships, and satellite data from Radarsat-2 to mention a few. These agencies know, however, that surveillance data is typically infrequent, irregular, and imprecise and might even have been faked by the perpetrators.
This year's Challenge problem focuses on one example problem: can we detect a cargo ship rendezvousing with another vessel at sea? Such rendezvous are seldom necessary to meet legitimate commercial objectives. The core problem is that the actual rendezvous will seldom be observed directly - it must be inferred or ruled out based on a time-series of positional information of ships coming form sparse, irregular and imprecise sensor data. This is a problem of model reconstruction from sparse data in one time and two space dimensions. Winning solutions must have a low false alarm rate, while still detecting the briefest of encounters.
The Challenge progressed in two stages. In the first stage, training dataset of simulated sensor data, along with the underlying ground-truth data, was provided to students. This allowed them to train their algorithm. In the next stage, a test dataset consisting of only sensor data was provided to students. Students were asked to run their trained algorithm on this dataset & submit results. In addition, students were asked to submit a description of their algorithm and code.
The team with the highest detection rate, lowest false alarm rate and best algorithm was deemed the winner.
Problem Statement
Undergraduate and graduate teams from across Canada, including from the University of Waterloo, University of British Columbia, Simon Fraser University and University of Toronto submitted solutions to the challenge. The undergraduate category winners were a team from the University of British Columbia. The graduate category winners were a team from Waterloo. The overall Challenge winners were an undergraduate team from University of British Columbia consisting of Adam Williams, and David Fagnan.